chunks per author: Spalding=13, Hunt=27, Linning=11, EthanSmith=27, Irving=65, Cooper=74, Rigdon=37, Cowdery=22, JSmith=9, Pratt=17, Phelps=31, Whitmer=15, KJV=60, JSrevelations=50, JSMoses=6, JSAbraham=2
same-author calibration: mean=0.587 p95=0.849

--- within-author register shift (one man, two registers) ---
JS letters -> D&C centroid: median=0.882
D&C -> JS letters centroid: median=0.905
(same-author p95 = 0.849; JS letters -> own centroid median = 0.691; D&C -> own centroid median = 0.683)

--- Book of Mormon chunks -> all corpora (median Delta) ---
  Linning        0.969
  JSrevelations  0.982
  Pratt          0.983
  JSMoses        0.989
  Phelps         0.999
  EthanSmith     1.006
  Rigdon         1.011
  KJV            1.040
  Whitmer        1.046
  Cowdery        1.049
  Hunt           1.052
  JSmith         1.055
  JSAbraham      1.070
  Spalding       1.099
  Irving         1.114
  Cooper         1.115
closed-set votes: JSrevelations:42, Linning:29, JSMoses:25, Hunt:9, Rigdon:8, KJV:3, JSmith:3, Phelps:2, EthanSmith:2, JSAbraham:1, Pratt:1, Cowdery:1
open-set: 56.3% of BoM chunks beyond same-author p95 from nearest corpus (median nearest=0.862)
BoM -> D&C: median=0.982, 19.8% of chunks within same-author range

--- discriminating the four dictated-scripture corpora ---
LOO-NN naive acc=0.98 (k=0.95) blocked acc=0.89 (k=0.77)
blocked per-corpus recall: Abraham=0.00(n=2), BoM=0.86(n=126), DC=0.98(n=50), Moses=1.00(n=6)
top confusions: BoM->DC:14, BoM->Moses:4, DC->BoM:1, Abraham->DC:1, Abraham->Moses:1
